mlsanityData to 7 Oct 2026

VVS: Accelerating Speculative Decoding for Visual Autoregressive Generation via Partial Verification Skipping

arXiv 2511.13587 ↗First posted 17 Nov 2025cs.CV cs.AIcited by 2

On the map: Efficient LLM inference and serving → Fast autoregressive visual generation

Abstract (arXiv)

Visual autoregressive (AR) generation models have demonstrated strong potential for image generation, yet their next-token-prediction paradigm introduces considerable inference latency. Although speculative decoding (SD) has been proven effective for accelerating visual AR models, its "draft one step, then verify one step" paradigm prevents a direct reduction in the number of forward passes, limiting its acceleration potential. Motivated by the interchangeability of visual tokens, we explore verification skipping in the SD process for the first time to explicitly cut the number of target model forward passes, thereby reducing inference latency. By analyzing the characteristics of the drafting stage, we observe that verification redundancy and stale feature reusability are key factors to maintain generation quality while improving speed for verification-free steps. Inspired by these two observations, we propose a novel SD framework VVS to accelerate visual AR model via partial verification skipping, which integrates three complementary modules: (1) a verification-free token selector with dynamic truncation, (2) token-level feature caching and reuse, and (3) fine-grained skipped step scheduling. Consequently, VVS reduces the number of target model forward passes by 2.8× relative to vanilla AR decoding while maintaining competitive generation quality, offering a superior speed-quality trade-off over conventional SD frameworks and revealing strong potential to reshape the SD paradigm. Our code is available at https://github.com/HyattDD/VVS.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Fast autoregressive visual generation, one of 12 directions in Efficient LLM inference and serving. Builds on 36 of its 40 references that are in the map. First posted on arXiv 17 Nov 2025.

Builds on

36 of 40 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025Grouped Speculative Decoding for Autoregressive Image Generation11 Aug
  2. 2024ZipAR: Parallel Auto-regressive Image Generation through Spatial Locality5 Dec
  3. 2024LANTERN: Accelerating Visual Autoregressive Models with Relaxed Speculative Decoding4 Oct
  4. 2024Accelerating Auto-regressive Text-to-Image Generation with Training-free Speculative Jacobi Decoding2 Oct

Other references · 32

  1. 2020Language Models are Few-Shot Learnerscited by 32,199
  2. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  3. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  4. 2021CLIPScore: A Reference-free Evaluation Metric for Image Captioningcited by 2,388
  5. 2022Fast Inference from Transformers via Speculative Decodingcited by 1,615
  6. 2022Diffusion Models: A Comprehensive Survey of Methods and Applicationscited by 1,107
  7. 2023Accelerating Large Language Model Decoding with Speculative Samplingcited by 975
  8. 2019Improved Precision and Recall Metric for Assessing Generative Modelscited by 925
  9. 2023Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesiscited by 847
  10. 2024Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Headscited by 821
  11. 2024Autoregressive Model Beats Diffusion: Llama for Scalable Image Generationcited by 760
  12. 2021Vector-quantized Image Modeling with Improved VQGANcited by 656
20 more
  1. 2024EAGLE: Speculative Sampling Requires Rethinking Feature Uncertaintycited by 564
  2. 2024EAGLE-2: Faster Inference of Language Models with Dynamic Draft Treescited by 362
  3. 2023SpecInfer: Accelerating Generative Large Language Model Serving with Tree-based Speculative Inference and Verificationcited by 359
  4. 2018Blockwise Parallel Decoding for Deep Autoregressive Modelscited by 318
  5. 2025EAGLE-3: Scaling up Inference Acceleration of Large Language Models via Training-Time Testcited by 318
  6. 2024Break the Sequential Dependency of LLM Inference Using Lookahead Decodingcited by 295
  7. 2024Unlocking Efficiency in Large Language Model Inference: A Comprehensive Survey of Speculative Decodingcited by 266
  8. 2022Speculative Decoding: Exploiting Speculative Execution for Accelerating Seq2seq Generationcited by 169
  9. 2024Lumina-mGPT: Illuminate Flexible Photorealistic Text-to-Image Generation with Multimodal Generative Pretrainingcited by 148
  10. 2023SpecTr: Fast Speculative Decoding via Optimal Transportcited by 129
  11. 2024ANOLE: An Open, Autoregressive, Native Large Multimodal Models for Interleaved Image-Text Generationcited by 112
  12. 2023SkipDecode: Autoregressive Skip Decoding with Batching and Caching for Efficient LLM Inferencecited by 89
  13. 2025Judge Decoding: Faster Speculative Sampling Requires Going Beyond Model Alignmentcited by 59
  14. 2025SP-VLA: A Joint Model Scheduling and Token Pruning Approach for VLA Model Accelerationcited by 56
  15. 2022Mixed-Precision Neural Network Quantization via Learned Layer-wise Importancecited by 53
  16. 2025Lumina-mGPT 2.0: Stand-Alone AutoRegressive Image Modelingcited by 35
  17. 2024Collaborative Decoding Makes Visual Auto-Regressive Modeling Efficientcited by 32
  18. 2024PRANCE: Joint Token-Optimization and Structural Channel-Pruning for Adaptive ViT Inferencecited by 10
  19. 2025Beyond A Single AI Cluster: A Survey of Decentralized LLM Trainingcited by 4
  20. 2024Staleness-Centric Optimizations for Parallel Diffusion MoE Inferencecited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026From Static Inference to Dynamic Interaction: A Survey of Streaming Large Language Modelscited by 3
  2. 2026Is Multimodal Speculative Decoding Ready for Diffusion-Based Parallel Drafting? A Survey and Empirical Diagnosiscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Speculative Coupled Decoding for Training-Free Lossless Acceleration of Autoregressive Visual Generation
  2. 2025LANTERN++: Enhancing Relaxed Speculative Decoding with Static Tree Drafting for Visual Auto-regressive Models
  3. 2024Continuous Speculative Decoding for Autoregressive Image Generation
  4. 2025SpecVLM: Fast Speculative Decoding in Vision-Language Models
  5. 2025From Sequential to Spatial: Reordering Autoregression for Efficient Visual Generation
  6. 2024RandAR: Decoder-only Autoregressive Visual Generation in Random Orders
  7. 2026Annealed Relaxation of Speculative Decoding for Faster Autoregressive Image Generation
  8. 2025Progress by Pieces: Test-Time Scaling for Autoregressive Image Generation